Project Details Statistical Models and Data Reductions to Estimate Standard Atomic Weights and Isotopic Ratios for the Elements, and to Evaluate the Associated Uncertainties

Project No.:
2019-024-1-200
Start Date:
10 December 2019
End Date:

Objective

To develop models and methods for assigning values to atomic weights and isotopic ratios using determinations, made on the same or different normal materials, which very often are mutually inconsistent. The project will expand, explore the boundaries of validity of, document, and apply systematically the multivariate meta-regression framework introduced by Meija & Possolo (2017, Metrologia 54; https://doi.org/10.1088/1681-7575/aa634d). The central goal is to produce a well-defined, peer-reviewed framework for value assignment to isotopic ratios and atomic weights, which the Commission on Isotopic Abundances and Atomic Weights (CIAAW) of the IUPAC may use as a reliable tool for the execution of its mission.

Description

The project aims to enact a paradigm change about how values are assigned to atomic weights. Up until recently, this was done mostly relying on a single measurement result deemed to be the best available. Meija & Possolo (2017) proposed statistical consensus building as a model-based alternative that accommodates and blends multiple determinations of the same isotopic ratios, in the same or different materials. Such multiple determinations typically are mutually inconsistent, in the sense that the variability of the corresponding measured values is appreciably and significantly larger than their individual, reported uncertainties. Such consensus-building approach is widely used in other areas of science and technology, in particular in medicine (where it is known as “meta-analysis”), and in metrology.

This project will apply the procedure originally proposed by Meija & Possolo (2017) systematically, to all available datasets of isotopic ratios relevant to the work of the CIAAW, to gauge the scope of its applicability, and the extent to which it produces substantively meaningful results. It is expected that this work will allow doing away with the application of arbitrary “expansion factors”, which were routinely applied in ad hoc attempts at reconciling mutually inconsistent data.

The project will also investigate other types of models that can be used for the same purpose, in particular Bayesian models for multvariate meta-analysis, and will produce demonstrations of their use to combine multiple measurements of several isotopic ratios simultaneously, this extending, in yet another direction, what Meija & Possolo (2017) did originally.

Progress

Page last updated 11 Dec 2019